Anchor-guided variance-aware reward modeling uses two response-level anchors to resolve non-identifiability in Gaussian models of pluralistic preferences, yielding provable identification, a joint training objective, and improved RLHF performance.
hub
Nature , volume=
11 Pith papers cite this work. Polarity classification is still indexing.
11
Pith papers citing it
hub tools
citation-role summary
background 1
citation-polarity summary
roles
background 1polarities
background 1representative citing papers
LOVER creates an unsupervised logic-regularized verifier that reaches 95% of supervised verifier performance on reasoning tasks across 10 datasets.
STAR-Teaming uses a Strategy-Response Multiplex Network inside a multi-agent framework to organize attack strategies into semantic communities, delivering higher attack success rates on LLMs at lower computational cost than prior methods.
Conflicting biomedical evidence triggers order-dependent prediction flips in RAG LLMs, and a new abstention score combining confidence with conflict detection raises selective accuracy by 7-33 points in the hardest conditions.
A latent mediation framework with sparse autoencoders enables non-additive token-level influence attribution in LLMs by learning orthogonal features and back-propagating attributions.
CodeClinic benchmark demonstrates that LLM-generated Python skill libraries from clinical guidelines enhance consistency and reduce token consumption by up to 40% compared to zero-shot approaches on MIMIC-IV based tasks.
LLMs exhibit sensitivity to small input changes and systematically under-recall rare and long-term side effects when listing radiation toxicities for breast cancer, with major gains from grounding in clinician-curated references.
Med-Gemini sets new records on 10 of 14 medical benchmarks including 91.1% on MedQA-USMLE, beats GPT-4V by 44.5% on multimodal tasks, and surpasses humans on medical text summarization.
LLMs assigned high or low status personas in multi-turn dialogues exhibit socio-cognitive effects including language coordination, pronoun patterns, persuasion success, and compliance with unsafe requests.
MedMSA framework retrieves knowledge via language models then builds formal probabilistic models to produce uncertainty-weighted differential diagnoses from symptoms.
A process-aware pipeline using logistic regression on prefix representations from 4,479 COVID-19 cases predicts ICU admission with AUC 0.906, improving from 0.642 early to 0.942 later in pathways.
citing papers explorer
-
CodeClinic: Evaluating Automation of Coding Skills for Clinical Reasoning Agents
CodeClinic benchmark demonstrates that LLM-generated Python skill libraries from clinical guidelines enhance consistency and reduce token consumption by up to 40% compared to zero-shot approaches on MIMIC-IV based tasks.
-
Capabilities of Gemini Models in Medicine
Med-Gemini sets new records on 10 of 14 medical benchmarks including 91.1% on MedQA-USMLE, beats GPT-4V by 44.5% on multimodal tasks, and surpasses humans on medical text summarization.
-
Medical Model Synthesis Architectures: A Case Study
MedMSA framework retrieves knowledge via language models then builds formal probabilistic models to produce uncertainty-weighted differential diagnoses from symptoms.